A fawn is trapped in a forest fire. It burns, takes several days to die, and is alone the whole time. No person ever learns of it. No lesson is drawn by anyone.

William Rowe put that example into a 1979 paper and it has been the center of the argument ever since, because it strips out everything the usual answers need. No free will was exercised. No character was built. Nobody grew.

The oldest version

Attributed to Epicurus, surviving first in Lactantius, and still the cleanest statement anyone has managed:

Is God willing to prevent evil, but not able? Then he is not omnipotent. Is he able, but not willing? Then he is malevolent. Is he both able and willing? Whence then is evil?

The force comes from the attributes doing the work. Omnipotence means able to prevent any particular evil. Omniscience means aware of it. Perfect goodness means motivated to stop it. Hold all three and it is hard to see how the Holocaust gets to happen.

This is the most-discussed single argument against theism in the Western tradition, and unlike most philosophical arguments, part of it has actually been resolved.

The part that got settled

For most of the twentieth century the argument was pressed in its logical form, sharpest in J. L. Mackie's 1955 paper: the existence of God and the existence of any evil at all are strictly incompatible, the way a square circle is incompatible. Not improbable — impossible.

Alvin Plantinga broke it in 1974, and the way he did it is worth understanding because it is so unlike how these arguments usually go.

He did not argue that God exists, or that any particular explanation of evil is true. He only had to show that God and evil are not contradictory, and for that he needed one possible scenario, however unlikely, in which they coexist. His candidate: it may be that every free creature God could have created would go wrong at least once. If that is so much as possible, then a world with free creatures and no evil may not have been available even to omnipotence, and the contradiction dissolves.

Most atheist philosophers accepted it. The logical problem of evil is generally regarded as defeated, by an argument its opponents conceded. That almost never happens in philosophy, and it is worth noticing when it does.

The part that did not

What replaced it is harder. Rowe's evidential version grants everything Plantinga established and asks a different question: not whether evil is compatible with God, but whether this much, of these kinds, is what you would expect if God existed.

Back to the fawn. Maybe some evils serve purposes. This one appears to serve none — it is invisible, it teaches nobody, and an omnipotent being could have prevented it at no cost to anything. Rowe calls that gratuitous.

The argument is probabilistic, and that is what makes it durable. Each individual case can be waved away with perhaps there is a reason we cannot see. The cumulative weight of millions of them is what does the damage. There is no single case to refute.

Lisbon, 1755: an earthquake on All Saints' Day killed tens of thousands while they were at Mass, and the churches collapsed on them. Voltaire wrote Candide four years later. It is the moment the problem stopped being a seminar exercise in Europe, and every serious treatment since has been written in its shadow.

Defense and theodicy are not the same move

This distinction is the hinge of the whole debate and it is constantly missed.

A defense shows that God and evil are logically compatible. It does not claim to be true. Plantinga's free will defense is a defense — he explicitly does not assert that transworld depravity is actual, only that it is possible. That is all a defense needs, and it is why the defense works against the logical argument and does nothing against the evidential one.

A theodicy is the stronger and more exposed claim: that it actually explains the evil we observe. It has to be true, not merely possible.

Almost every unsatisfying conversation about this is someone offering a defense where a theodicy was wanted, or attacking a defense for failing to be a theodicy. What the theodicies are and what each of them costs is its own subject.

The response that may prove too much

Skeptical theism is the most serious reply to Rowe. Stephen Wykstra's version: you are not well placed to judge that an evil is pointless. An omniscient being would have reasons you could not grasp, and from where you stand, an evil with no purpose and an evil whose purpose you cannot perceive look exactly alike. So the move from I see no good reason for this to there is no good reason for this is not licensed.

This is a real objection and it has a real cost. If your judgment about whether suffering is pointless is that unreliable, what happens to every other moral judgment you make without a God's-eye view? Critics — including Rowe — argue it buys a defense against the evidential argument at the price of general moral skepticism.

Whether that is principled epistemic humility or a convenient blindfold is the live dispute.

Why it is worth your time

This is the argument that moves people. More people have left a religion over a dead child than over the cosmological argument, and no amount of technical philosophy of religion changes that.

Which is why the last objection on the page is the most serious one. A tradition running through Levinas and Ricoeur argues that a theodicy is the wrong kind of response — that explaining why a particular child's suffering was permissible, even successfully, is a moral failure rather than an intellectual achievement. On this view, the correct answer to the problem of evil is not a better theodicy but a refusal to produce one.

Whatever you conclude, the argument is the best available demonstration that a position can be logically defensible and still not survive contact with what it is about.

Further reading